CN105654437A - Enhancement method for low-illumination image - Google Patents

Abstract

The invention discloses an enhancement method for a low-illumination image. The enhancement method comprises the steps that the low-illumination image to be processed is acquired and then the low-illumination image to be processed is transferred to an HSV color space from an RGB color space, and a chroma component, a saturation component and a brightness component are acquired; the brightness component is decomposed into a reflection component and an irradiation component by adopting an alternating minimization method based on a Retinex algorithm; the irradiation component and the reflection component are respectively enhanced and then synthesized into an enhanced brightness component; the saturation component is self-adaptively adjusted and then an enhanced saturation component is obtained; the chroma component, the enhanced brightness component and the enhanced saturation component are synthesized into a new HSV image; and the obtained new HSV image is converted into an RGB image, and an enhanced image is obtained through white balance processing. Definition of the low-illumination image can be greatly enhanced, and details are enabled to be reproduced so that the method is high in applicability and high in robustness and can be widely applied to the field of image processing.

Description

A kind of Enhancement Method to low illumination image

Technical field

The present invention relates to image processing field, particularly relate to a kind of Enhancement Method to low illumination image.

Background technology

Explanation of nouns:

Retinex algorithm: Retinex is " Retina " (retina) and the abbreviation of " Cortex " (pallium). Retine algorithm is a kind of algorithm for image enhancement based on human visual system being based upon on scientific experiment and scientific analysis basis, can compress in dynamicrange, edge strengthen and color constancy three in reach balance, various dissimilar image is carried out the enhancing of adaptivity;

RGB: comprise red component R, green component G, blue component B tri-components.

HSV: comprise chromatic(ity)component H, saturation ratio component S, brightness component V tri-components.

Computer vision monitoring equipment is under the low light conditions such as indoor, night, owing to the illumination of non-natural light source is insufficient, so the reflected light of target surface is more weak, cause the insufficient light entering imaging sensor, cause that the deteriroation of image quality gathered at night is serious, image can identification very low, and containing a large amount of noise, to such an extent as to the details being difficult in resolution image, greatly reducing the using value of image, this kind of image is referred to as low illumination image.

To the increased quality of low illumination image, being one of the research focus that current picture quality promotes field, especially at the computer vision field such as Urban traffic, monitor video, the increased quality of low illumination image is significant. At present image mainly directly being carried out under RGB color by the Enhancement Method of low illumination image enhancement process such as MSRCR algorithm, this kind of processing mode easily produces color and loses, and does not meet the visual signature of people's eye.

Summary of the invention

In order to solve above-mentioned technical problem, it is an object of the invention to provide a kind of Enhancement Method to low illumination image.

The technical solution adopted for the present invention to solve the technical problems is:

To an Enhancement Method for low illumination image, comprising:

S1, obtain pending low illumination image after, it is transformed into hsv color space from RGB color, obtains chromatic(ity)component, saturation ratio component and brightness component;

S2, based on Retinex algorithm, alternately minimization method is adopted brightness component to be decomposed into reflection component and irradiates component;

S3, respectively to after irradiating component and reflection component carrying out enhancement process, synthesis strengthen after brightness component;

S4, saturation ratio component is done self-adaptative adjustment process after obtain strengthen after saturation ratio component;

S5, by chromatic(ity)component, strengthen after brightness component and strengthen after saturation ratio component synthesize new HSV image;

S6, the new HSV image obtained is converted into RGB image, and after carrying out white balance process, obtains the image after strengthening.

Further, described step S2, it is specially:

Build following objective function, and obtain optimum irradiation component and reflection component according to this objective function:

$\begin{array}{ccc}\underset{R,L}{\mathrm{argmin}}\left|\right|RL-V|{|}_2^2+\mathrm{\α}||\▿L|{|}_2^2+\mathrm{\β}\left|\right|\▿R|{|}_0+\mathrm{\λ}||L-I_{bright}|{|}_2^2& s.t.& 0\≤R\≤1,I\≤L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component,Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor.

Further, described step S2, comprising:

S21, build following objective function:

$\begin{array}{ccc}\underset{R,L}{\mathrm{argmin}}\left|\right|RL-V|{|}_2^2+\mathrm{\α}||\▿L|{|}_2^2+\mathrm{\β}\left|\right|\▿R|{|}_0+\mathrm{\λ}||L-I_{bright}|{|}_2^2& s.t.& 0\≤R\≤1,I\≤L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component,Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor;

S22, structureEquivalent expression: $\left|\right|\▿R|{|}_0=C\left(R\right)=\#\{p\left|\right|\∂{\mathrm{xR}}_p|+|\∂{\mathrm{yR}}_p|\≠0\},$ Wherein p represents and satisfies conditionThe number of pixel;

S23, obtain according to following formulaSubsidiary variable:

$(h_p,v_p)=\left\{\begin{array}{ccc}(0,0)& if& {(\∂{\mathrm{xR}}_p)}^2+{(\∂{\mathrm{yR}}_p)}^2\≤\mathrm{\β}/\mathrm{\γ}\\ (\∂{\mathrm{xR}}_p,\∂{\mathrm{yR}}_p)& if& others\end{array}\right.$

Wherein, h_pRepresentSubsidiary variable, v_pRepresentSubsidiary variable, (h_p,v_p) meet $\underset{h_p,v_p}{min}\{\underset{p}{\Σ}({\left(\∂{\mathrm{xR}}_p-h_p\right)}^2+{\left(\∂{\mathrm{yR}}_p-v_p\right)}^2)+\frac{\mathrm{\β}}{{\mathrm{\γ}}_i}C(h,v)\},$ ��_iRepresent iteration variable, ��₁=2 ��=0.05 and ��_i+1=2* ��_i;

S24, using brightness component by gauss low frequency filter filtered value as irradiation component initial value;

S25, the reflection component obtaining optimum according to objective function are as follows:

$R=f^{-1}\frac{f(V/L_0)+{\mathrm{\γ}}_i(f{\left(\∂x\right)}^{*}\×f(h)+f{\left(\∂y\right)}^{*}\×f(v\left)\right)}{f\left(1\right)+{\mathrm{\γ}}_i(f{\left(\∂x\right)}^{*}\×f(\∂x)+f{\left(\∂y\right)}^{*}\×f(\∂y\left)\right)}$

Wherein, L₀Representing the initial value irradiating component, f represents FFT, f^-1Represent FFT inverse transformation, f ()^*Hetero conjugation after expression FFT, h representsSubsidiary variable;

S26, according to following formula to reflection variable correct: R=min (max (R, 0), 1);

The value of the reflection component after S27, maintenance rectification is constant, obtains optimum irradiation component according to objective function as follows:

$L=f^{-1}\frac{f(V/R+{\mathrm{\λI}}_{bright})}{f(1+\mathrm{\λ})+\mathrm{\α}(f{\left(\∂x\right)}^{*}\×f(\∂x)+f{\left(\∂y\right)}^{*}\×f(\∂y\left)\right)}$

S28, according to following formula to irradiation variable correct: L=max (L, V);

S29, judge iteration variable ��_iWhether it is greater than predetermined threshold value, if then terminating, otherwise after iteration number of times is added 1, using the irradiation component after rectification as the initial value irradiating component, returning and performing step S25.

Further, described step S3, comprising:

S31, irradiation component is carried out gamma correction, and after carrying out adaptive histogram equalization process, obtain the irradiation component after strengthening;

S32, reflection component is carried out obtaining after part filter operation the reflection component after strengthening;

S33, by strengthen after irradiation component and strengthen after reflection component synthesis strengthen after brightness component.

Further, irradiation component carrying out described in described step S31 the step of gamma correction, it is specially:

According to following formula, irradiation component is carried out gamma correction:

$L_g=L^{\frac{1}{\mathrm{\γ}}},\mathrm{\γ}=2.2$

In upper formula, L represents irradiation component, L_gRepresent the irradiation component after gamma correction.

Further, described step S32, it is specially:

According to following formula, after reflection component is carried out part filter operation, obtain the reflection component after strengthening:

$R_F(x,y)=\frac{1}{K}\×\underset{(i,j)\∈\mathrm{\Ω}}{\Σ}{w}_s(i,j)w_r(i,j)R(i,j)$

In upper formula, R_F(x, y) representing that the reflection after the reflection component of pixel (x, y) is carried out part filter operation divides value, �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window, R (i, j) pixel (i is represented, j) reflection divides value, and K represents normalization method constant, w_s(i, j) represents the spatial domain weight of pixel (i, j), w_r(i, j) represents the codomain weight of pixel (i, j), and K, w_s(i,j)��w_r(i, j) meets the following conditions:

$\left\{\begin{array}{c}K=\underset{(i,j)\∈\mathrm{\Ω}}{\Σ}{w}_s(i,j)w_r(i,j)\\ w_s(i,j)=\exp (-\frac{|i-x{|}^2+|j-y{|}^2}{2{\mathrm{\σ}}_s^2})\\ w_r(i,j)=\underset{m=x}{\overset{i-1}{\mathrm{\Σ}}}{w}_r(m,y;m+1,y)\×\underset{n=y}{\overset{j-1}{\mathrm{\Σ}}}{w}_r(i,n;i,n+1)=\exp (-\frac{\underset{m=x}{\overset{i-1}{\mathrm{\Σ}}}|R_{m,y}-R_{m+1,y}{|}^2+\underset{n=y}{\overset{i-1}{\mathrm{\Σ}}}|R_{i,n}-R_{i,n+1}{|}^2}{2{\mathrm{\σ}}_r^2})\end{array}\right.$

Wherein, ��_SRepresent spatial domain standard deviation, ��_rRepresent codomain standard deviation, R_m,yRepresent the intensity level of pixel (m, y), R_i,nRepresent the intensity level of pixel (i, n).

Further, described step S4, it is specially:

According to following formula, saturation ratio component is done after self-adaptative adjustment processes and obtains the saturation ratio component after strengthening:

S'=S+t �� (V'-V) �� ��

In upper formula, S' represents the saturation ratio component after enhancing, and S represents saturation ratio component, and t is constant, and V represents brightness component, and V' represents the brightness component after enhancing, and �� represents regulation coefficient, and for each pixel (x, y), regulation coefficient is:

$\mathrm{\μ}(x,y)=\underset{(i,j)\∈\mathrm{\Ω}}{\Σ}\left(\right|\left(V\left(i,j\right)-\stackrel{\‾}{V_{\mathrm{\Ω}}}\left(i,j\right)\right)|\×|\left(S\left(i,j\right)-\stackrel{\‾}{S_{\mathrm{\Ω}}}\left(i,j\right)\right)\left|\right)/\sqrt{{\mathrm{\δ}}_V(x,y)\×{\mathrm{\δ}}_S(x,y)}$

Wherein, (x, y) represents pixel position, and �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window ��, V (i, j) brightness value of pixel (i, j) is representedRepresenting the luminance mean value in neighborhood window ��, S (i, j) represents the intensity value of pixel (i, j),Represent the saturation ratio average in neighborhood window ��, ��_V(x, y) represents the brightness variance of pixel (x, y), ��_S(x, y) represents the saturation ratio variance of (x, the y) of pixel.

Further, carry out the step of white balance process described in described step S6, comprising:

S61, the color average calculating R, G, B triple channel obtaining the RGB image after transforming, and total average of RGB channel is calculated according to following formula:

K_ave=(R_ave+G_ave+B_ave)/3

In upper formula, K_aveRepresent total average of RGB channel, R_ave��G_aveB_aveRepresent the color average of R, G, B triple channel respectively;

S62, judge following formula and whether set up, if then directly terminating, otherwise, perform step S63:

$\frac{max(R_{ave},G_{ave},B_{ave})-min(R_{ave},G_{ave},B_{ave})}{K_{ave}}<0.1$

S63, according to following formula, R, G, B color component is carried out white balance process:

$\left\{\begin{array}{c}{R}_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/R_{ave})\&times;R_{channel}\\ G_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/G_{ave})\&times;G_{channel}\\ B_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/B_{ave})\&times;B_{channel}\end{array}\right.$

In upper formula, R_mod��G_mod��B_modRepresent the color value of R, G, B triple channel after white balance process respectively, R_channel��G_channel��B_channelRepresent the gray-scale value of the front R of white balance process, G, channel B respectively.

The invention has the beneficial effects as follows: a kind of Enhancement Method to low illumination image of the present invention, comprising: after obtaining pending low illumination image, from RGB color, it is transformed into hsv color space, obtain chromatic(ity)component, saturation ratio component and brightness component; Based on Retinex algorithm, alternately minimization method is adopted brightness component to be decomposed into reflection component and irradiates component; After respectively irradiation component and reflection component being carried out enhancement process, the brightness component after synthesis enhancing; Saturation ratio component is done after self-adaptative adjustment processes and obtain the saturation ratio component after strengthening; Brightness component after chromatic(ity)component, enhancing and the saturation ratio component after enhancing are synthesized new HSV image; The new HSV image obtained is converted into RGB image, and after carrying out white balance process, obtains the image after strengthening. Present method can greatly promote the sharpness of low illumination image, and details can be reproduced, and substantially can not produce color loss, more meets the visual signature of people's eye, and suitability is strong, robustness height.

Accompanying drawing explanation

Below in conjunction with drawings and Examples, the invention will be further described.

Fig. 1 is the schematic flow sheet of a kind of Enhancement Method to low illumination image of the present invention;

Fig. 2 is the pending low illumination image in specific embodiments of the invention;

Fig. 3 is the result obtained after adopting tradition Enhancement Method that the low illumination image of Fig. 2 carries out enhancement process;

Fig. 4 is the result obtained after the low illumination image of Fig. 2 carries out in specific embodiments of the invention enhancement process.

Embodiment

With reference to Fig. 1, the present invention provides a kind of Enhancement Method to low illumination image, comprising:

S1, obtain pending low illumination image after, it is transformed into hsv color space from RGB color, obtains chromatic(ity)component, saturation ratio component and brightness component;

S2, based on Retinex algorithm, alternately minimization method is adopted brightness component to be decomposed into reflection component and irradiates component;

S3, respectively to after irradiating component and reflection component carrying out enhancement process, synthesis strengthen after brightness component;

S4, saturation ratio component is done self-adaptative adjustment process after obtain strengthen after saturation ratio component;

S5, by chromatic(ity)component, strengthen after brightness component and strengthen after saturation ratio component synthesize new HSV image;

S6, the new HSV image obtained is converted into RGB image, and after carrying out white balance process, obtains the image after strengthening.

Being further used as preferred embodiment, described step S2, it is specially:

Build following objective function, and obtain optimum irradiation component and reflection component according to this objective function:

$\begin{array}{ccc}\underset{R,L}{\mathrm{argmin}}\left|\right|RL-V|{|}_2^2+\mathrm{\&alpha;}||\&dtri;L|{|}_2^2+\mathrm{\&beta;}\left|\right|\&dtri;R|{|}_0+\mathrm{\&lambda;}||L-I_{bright}|{|}_2^2& s.t.& 0\&le;R\&le;1,I\&le;L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component,Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor.

Being further used as preferred embodiment, described step S2, comprising:

S21, build following objective function:

$\begin{array}{ccc}\underset{R,L}{\mathrm{argmin}}\left|\right|RL-V|{|}_2^2+\mathrm{\&alpha;}||\&dtri;L|{|}_2^2+\mathrm{\&beta;}\left|\right|\&dtri;R|{|}_0+\mathrm{\&lambda;}||L-I_{bright}|{|}_2^2& s.t.& 0\&le;R\&le;1,I\&le;L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component,Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor;

S22, structureEquivalent expression: $\left|\right|\&dtri;R|{|}_0=C\left(R\right)=\#\{p\left|\right|\&part;{\mathrm{xR}}_p|+|\&part;{\mathrm{yR}}_p|\&NotEqual;0\},$ Wherein p represents and satisfies conditionThe number of pixel;

S23, obtain according to following formulaSubsidiary variable:

$(h_p,v_p)=\left\{\begin{array}{ccc}(0,0)& if& {(\&part;{\mathrm{xR}}_p)}^2+{(\&part;{\mathrm{yR}}_p)}^2\&le;\mathrm{\&beta;}/\mathrm{\&gamma;}\\ (\&part;{\mathrm{xR}}_p,\&part;{\mathrm{yR}}_p)& if& others\end{array}\right.$

Wherein, h_pRepresentSubsidiary variable, v_pRepresentSubsidiary variable, (h_p,v_p) meet $\underset{h_p,v_p}{min}\{\underset{p}{\&Sigma;}({\left(\&part;{\mathrm{xR}}_p-h_p\right)}^2+{\left(\&part;{\mathrm{yR}}_p-v_p\right)}^2)+\frac{\mathrm{\&beta;}}{{\mathrm{\&gamma;}}_i}C(h,v)\},$ ��_iRepresent iteration variable, ��₁=2 ��=0.05 and ��_i+1=2* ��_i;

S24, using brightness component by gauss low frequency filter filtered value as irradiation component initial value;

S25, the reflection component obtaining optimum according to objective function are as follows:

$R=f^{-1}\frac{f(V/L_0)+{\mathrm{\&gamma;}}_i(f{\left(\&part;x\right)}^{*}\&times;f(h)+f{\left(\&part;y\right)}^{*}\&times;f(v\left)\right)}{f\left(1\right)+{\mathrm{\&gamma;}}_i(f{\left(\&part;x\right)}^{*}\&times;f(\&part;x)+f{\left(\&part;y\right)}^{*}\&times;f(\&part;y\left)\right)}$

Wherein, L₀Representing the initial value irradiating component, f represents FFT, f^-1Represent FFT inverse transformation, f ()^*Hetero conjugation after expression FFT, h representsSubsidiary variable;

S26, according to following formula to reflection variable correct: R=min (max (R, 0), 1);

The value of the reflection component after S27, maintenance rectification is constant, obtains optimum irradiation component according to objective function as follows:

$L=f^{-1}\frac{f(V/R+{\mathrm{\&lambda;I}}_{bright})}{f(1+\mathrm{\&lambda;})+\mathrm{\&alpha;}(f{\left(\&part;x\right)}^{*}\&times;f(\&part;x)+f{\left(\&part;y\right)}^{*}\&times;f(\&part;y\left)\right)}$

S28, according to following formula to irradiation variable correct: L=max (L, V);

S29, judge whether iteration number of times is greater than predetermined threshold value, if then terminating, otherwise after iteration number of times is added 1, using the irradiation component after correcting as the initial value irradiating component, returning and performing step S25.

Being further used as preferred embodiment, described step S3, comprising:

S31, irradiation component is carried out gamma correction, and after carrying out adaptive histogram equalization process, obtain the irradiation component after strengthening;

S32, reflection component is carried out obtaining after part filter operation the reflection component after strengthening;

S33, by strengthen after irradiation component and strengthen after reflection component synthesis strengthen after brightness component.

Being further used as preferred embodiment, irradiation component carries out described in described step S31 the step of gamma correction, it is specially:

According to following formula, irradiation component is carried out gamma correction:

$L_g=L^{\frac{1}{\mathrm{\&gamma;}}},\mathrm{\&gamma;}=2.2$

In upper formula, L represents irradiation component, L_gRepresent the irradiation component after gamma correction.

Being further used as preferred embodiment, described step S32, it is specially:

According to following formula, after reflection component is carried out part filter operation, obtain the reflection component after strengthening:

$R_F(x,y)=\frac{1}{K}\&times;\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}{w}_s(i,j)w_r(i,j)R(i,j)$

In upper formula, R_F(x, y) representing that the reflection after the reflection component of pixel (x, y) is carried out part filter operation divides value, �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window, R (i, j) pixel (i is represented, j) reflection divides value, and K represents normalization method constant, w_s(i, j) represents the spatial domain weight of pixel (i, j), w_r(i, j) represents the codomain weight of pixel (i, j), and K, w_s(i,j)��w_r(i, j) meets the following conditions:

$\left\{\begin{array}{c}K=\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}{w}_s(i,j)w_r(i,j)\\ w_s(i,j)=\exp (-\frac{|i-x{|}^2+|j-y{|}^2}{2{\mathrm{\&sigma;}}_s^2})\\ w_r(i,j)=\underset{m=x}{\overset{i-1}{\mathrm{\&Sigma;}}}{w}_r(m,y;m+1,y)\&times;\underset{n=y}{\overset{j-1}{\mathrm{\&Sigma;}}}{w}_r(i,n;i,n+1)=\exp (-\frac{\underset{m=x}{\overset{i-1}{\mathrm{\&Sigma;}}}|R_{m,y}-R_{m+1,y}{|}^2+\underset{n=y}{\overset{j-1}{\mathrm{\&Sigma;}}}|R_{i,n}-R_{i,n+1}{|}^2}{2{\mathrm{\&sigma;}}_r^2})\end{array}\right.$

Wherein, ��_SRepresent spatial domain standard deviation, ��_rRepresent codomain standard deviation, R_m,yRepresent the intensity level of pixel (m, y), R_i,nRepresent the intensity level of pixel (i, n).

Being further used as preferred embodiment, described step S4, it is specially:

According to following formula, saturation ratio component is done after self-adaptative adjustment processes and obtains the saturation ratio component after strengthening:

S'=S+t �� (V'-V) �� ��

In upper formula, S' represents the saturation ratio component after enhancing, and S represents saturation ratio component, and t is constant, and V represents brightness component, and V' represents the brightness component after enhancing, and �� represents regulation coefficient, and for each pixel (x, y), regulation coefficient is:

$\mathrm{\&mu;}(x,y)=\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}\left(\right|\left(V\left(i,j\right)-\stackrel{\&OverBar;}{V_{\mathrm{\&Omega;}}}\left(i,j\right)\right)|\&times;|\left(S\left(i,j\right)-\stackrel{\&OverBar;}{S_{\mathrm{\&Omega;}}}\left(i,j\right)\right)\left|\right)/\sqrt{{\mathrm{\&delta;}}_V(x,y)\&times;{\mathrm{\&delta;}}_S(x,y)}$

Wherein, (x, y) represents pixel position, and �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window ��, V (i, j) brightness value of pixel (i, j) is representedRepresenting the luminance mean value in neighborhood window ��, S (i, j) represents the intensity value of pixel (i, j),Represent the saturation ratio average in neighborhood window ��, ��_V(x, y) represents the brightness variance of pixel (x, y), ��_S(x, y) represents the saturation ratio variance of (x, the y) of pixel.

It is further used as preferred embodiment, carries out the step of white balance process described in described step S6, comprising:

S61, the color average calculating R, G, B triple channel obtaining the RGB image after transforming, and total average of RGB channel is calculated according to following formula:

K_ave=(R_ave+G_ave+B_ave)/3

In upper formula, K_aveRepresent total average of RGB channel, R_ave��G_aveB_aveRepresent the color average of R, G, B triple channel respectively;

S62, judge following formula and whether set up, if then directly terminating, otherwise, perform step S63:

$\frac{max(R_{ave},G_{ave},B_{ave})-min(R_{ave},G_{ave},B_{ave})}{K_{ave}}<0.1$

S63, according to following formula, R, G, B color component is carried out white balance process:

$\left\{\begin{array}{c}{R}_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/R_{ave})\&times;R_{channel}\\ G_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/G_{ave})\&times;G_{channel}\\ B_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/B_{ave})\&times;B_{channel}\end{array}\right.$

In upper formula, R_mod��G_mod��B_modRepresent the color value of R, G, B triple channel after white balance process respectively, R_channel��G_channel��B_channelRepresent the gray-scale value of the front R of white balance process, G, channel B respectively.

Below in conjunction with a specific embodiment, the present invention is elaborated.

With reference to Fig. 1, a kind of Enhancement Method to low illumination image, comprising:

S1, obtain pending low illumination image after, it is transformed into hsv color space from RGB color, obtains chromatic(ity)component, saturation ratio component and brightness component;

S2, based on Retinex algorithm, adopt alternately minimization method brightness component is decomposed into reflection component and irradiates component, it be specially: build following objective function, and obtain optimum irradiation component and reflection component according to this objective function:

$\begin{array}{ccc}\underset{R,L}{\mathrm{argmin}}\left|\right|RL-V|{|}_2^2+\mathrm{\&alpha;}||\&dtri;L|{|}_2^2+\mathrm{\&beta;}\left|\right|\&dtri;R|{|}_0+\mathrm{\&lambda;}||L-I_{bright}|{|}_2^2& s.t.& 0\&le;R\&le;1,I\&le;L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component,Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor.

In more detail, step S2 comprises S21��S27:

S21, build following objective function:

$\begin{array}{ccc}\underset{R,L}{\mathrm{argmin}}\left|\right|RL-V|{|}_2^2+\mathrm{\&alpha;}||\&dtri;L|{|}_2^2+\mathrm{\&beta;}\left|\right|\&dtri;R|{|}_0+\mathrm{\&lambda;}||L-I_{bright}|{|}_2^2& s.t.& 0\&le;R\&le;1,I\&le;L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component,Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor;

S22, structureEquivalent expression: $\left|\right|\&dtri;R|{|}_0=C\left(R\right)=\#\{p\left|\right|\&part;{\mathrm{xR}}_p|+|\&part;{\mathrm{yR}}_p|\&NotEqual;0\},$ Wherein p represents and satisfies conditionThe number of pixel; The operation that gradient is not the pixel P of 0 is asked in C (R) expression, and the gradient of P point is

S23, obtain according to following formulaSubsidiary variable:

$(h_p,v_p)=\left\{\begin{array}{ccc}(0,0)& if& {(\&part;{\mathrm{xR}}_p)}^2+{(\&part;{\mathrm{yR}}_p)}^2\&le;\mathrm{\&beta;}/\mathrm{\&gamma;}\\ (\&part;{\mathrm{xR}}_p,\&part;{\mathrm{yR}}_p)& if& others\end{array}\right.$

Wherein, in this formula, if represents when meeting what condition, when ifothers represents other situation; Wherein, h_pRepresentSubsidiary variable, v_pRepresentSubsidiary variable, (h_p,v_p) meet $\underset{h_p,v_p}{min}\{\underset{p}{\&Sigma;}({\left(\&part;{\mathrm{xR}}_p-h_p\right)}^2+{\left(\&part;{\mathrm{yR}}_p-v_p\right)}^2)+\frac{\mathrm{\&beta;}}{{\mathrm{\&gamma;}}_i}C(h,v)\},$ ��_iRepresent iteration variable, ��₁After=2 ��=0.05 and every time iteration, iteration variable is changed to ��_i+1=2* ��_i;In step S29, predetermined threshold value is set to 0.5 i.e. successive ignition until ��_i> 0.5; Pass through �� herein_iControl iteration number of times about 5 times;

Due to the element number that L0 norm is non-zero in statistical vector, cannot directly carry out derivative operation, therefore by (h in upper formula_p,v_p) expression formula substitute into the objective function in step S21, former objective function is changed into:

$\underset{R,L}{\mathrm{argmin}}\left|\right|RL-V|{|}_2^2+\mathrm{\&alpha;}||\&dtri;L|{|}_2^2+\mathrm{\&beta;}({\left(\&part;xR-h\right)}^2+{\left(\&part;xR-v\right)}^2)+\mathrm{\&tau;}C(h,v)+\mathrm{\&lambda;}\left|\right|L-I_{bright}|{|}_2^2$

Wherein,H(|h_p|+|v_p|) it is a binary function, when | h_p|+|v_p| when �� 0, return 1; Other situations, return 0;

S24, using brightness component by gauss low frequency filter filtered value as irradiation component initial value;

S25, the reflection component obtaining optimum according to objective function are as follows:

$R=f^{-1}\frac{f(V/L_0)+{\mathrm{\&gamma;}}_i(f{\left(\&part;x\right)}^{*}\&times;f(h)+f{\left(\&part;y\right)}^{*}\&times;f(v\left)\right)}{f\left(1\right)+{\mathrm{\&gamma;}}_i(f{\left(\&part;x\right)}^{*}\&times;f(\&part;x)+f{\left(\&part;y\right)}^{*}\&times;f(\&part;y\left)\right)}$

Wherein, L₀Representing the initial value irradiating component, f represents FFT, f^-1Represent FFT inverse transformation, f ()^*Hetero conjugation after expression FFT, h representsSubsidiary variable;

S26, according to following formula to reflection variable correct: R=min (max (R, 0), 1);

The value of the reflection component after S27, maintenance rectification constant (being equivalent to the reflection component after using rectification here as asking for reflection component initial value when irradiating component), the irradiation component obtaining optimum according to objective function is as follows:

$L=f^{-1}\frac{f(V/R+{\mathrm{\&lambda;I}}_{bright})}{f(1+\mathrm{\&lambda;})+\mathrm{\&alpha;}(f{\left(\&part;x\right)}^{*}\&times;f(\&part;x)+f{\left(\&part;y\right)}^{*}\&times;f(\&part;y\left)\right)}$

S28, according to following formula to irradiation variable correct: L=max (L, V);

S29, judge iteration variable ��_iWhether it is greater than predetermined threshold value, if then terminating, otherwise after iteration number of times is added 1, using the irradiation component after rectification as the initial value irradiating component, returning and performing step S25. In the present embodiment, it is preferable that predetermined threshold value is set to 0.5, after about 5 iteration, �� can be met_i> 0.5, it is possible to obtain the irradiation component L tended towards stability and reflection components R. The present embodiment passes through ��_iControl iteration number of times, it is preferable that control iteration number of times is about 5 times.

S3, respectively to after irradiating component and reflection component carrying out enhancement process, synthesis strengthen after brightness component; Step S3 comprises S31��S33:

S31, according to following formula to irradiation component carry out gamma correction, obtain gamma correct after irradiation component L_g, and after it is carried out adaptive histogram equalization process, obtain the irradiation component L' after strengthening:

$L_g=L^{\frac{1}{\mathrm{\&gamma;}}},\mathrm{\&gamma;}=2.2$

In upper formula, L represents irradiation component, L_gRepresent the irradiation component after gamma correction;

Due to the even factor of uneven illumination, there is obvious brightness light and shade region in the irradiation component L obtained, it is necessary to it is carried out gamma correction. Finally, the irradiation component L':L'=CLAHE (L after strengthening is obtained after adopting the histogram equalization method of prior art to carry out histogram equalization processing_g), CLAHE represents self-adapting histogram equilibrium function.

S32, reflection component is carried out obtaining after part filter operation the reflection component after strengthening; The reflection component of image changes high-frequency information faster in image, the inherent character of object on image can be reflected, extract in the process obtaining reflection component adopting Retinex theory, owing to noise is usually distributed in high frequency, noise has been exaggerated, therefore this step needs to carry out part filter and realizes noise reduction process, specific as follows:

According to following formula, after reflection component is carried out part filter operation, obtain the reflection component after strengthening:

$R_F(x,y)=\frac{1}{K}\&times;\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}{w}_s(i,j)w_r(i,j)R(i,j)$

In upper formula, R_F(x, y) representing that the reflection after the reflection component of pixel (x, y) is carried out part filter operation divides value, �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window, R (i, j) pixel (i is represented, j) reflection divides value, and K represents normalization method constant, w_s(i, j) represents the spatial domain weight of pixel (i, j), w_r(i, j) represents the codomain weight of pixel (i, j), and K, w_s(i,j)��w_r(i, j) meets the following conditions:

$\left\{\begin{array}{c}K=\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}{w}_s(i,j)w_r(i,j)\\ w_s(i,j)=\exp (-\frac{|i-x{|}^2+|j-y{|}^2}{2{\mathrm{\&sigma;}}_s^2})\\ w_r(i,j)=\underset{m=x}{\overset{i-1}{\mathrm{\&Sigma;}}}{w}_r(m,y;m+1,y)\&times;\underset{n=y}{\overset{j-1}{\mathrm{\&Sigma;}}}{w}_r(i,n;i,n+1)=\exp (-\frac{\underset{m=x}{\overset{i-1}{\mathrm{\&Sigma;}}}|R_{m,y}-R_{m+1,y}{|}^2+\underset{n=y}{\overset{j-1}{\mathrm{\&Sigma;}}}|R_{i,n}-R_{i,n+1}{|}^2}{2{\mathrm{\&sigma;}}_r^2})\end{array}\right.$

Wherein, ��_SRepresent spatial domain standard deviation, ��_rRepresent codomain standard deviation, R_m,yRepresent the intensity level of pixel (m, y), R_i,nRepresent the intensity level of pixel (i, n).

S33, by strengthen after irradiation component and strengthen after reflection component synthesis strengthen after brightness component.

S4, saturation ratio component is done self-adaptative adjustment process after obtain strengthen after saturation ratio component, specific as follows:

According to following formula, saturation ratio component is done after self-adaptative adjustment processes and obtains the saturation ratio component after strengthening:

S'=S+t �� (V'-V) �� ��

In upper formula, S' represents the saturation ratio component after enhancing, and S represents saturation ratio component, and t is constant, and V represents brightness component, and V' represents the brightness component after enhancing, and �� represents regulation coefficient, and for each pixel (x, y), regulation coefficient is:

$\mathrm{\&mu;}(x,y)=\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}\left(\right|\left(V\left(i,j\right)-\stackrel{\&OverBar;}{V_{\mathrm{\&Omega;}}}\left(i,j\right)\right)|\&times;|\left(S\left(i,j\right)-\stackrel{\&OverBar;}{S_{\mathrm{\&Omega;}}}\left(i,j\right)\right)\left|\right)/\sqrt{{\mathrm{\&delta;}}_V(x,y)\&times;{\mathrm{\&delta;}}_S(x,y)}$

Wherein, (x, y) represents pixel position, and �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window ��, V (i, j) brightness value of pixel (i, j) is representedRepresenting the luminance mean value in neighborhood window ��, S (i, j) represents the intensity value of pixel (i, j),Represent the saturation ratio average in neighborhood window ��, ��_V(x, y) represents the brightness variance of pixel (x, y), ��_S(x, y) represents the saturation ratio variance of (x, the y) of pixel.

In S5, treating processes, chromatic(ity)component remains unchanged, and in this step, the brightness component after chromatic(ity)component, enhancing and the saturation ratio component after enhancing is synthesized new HSV image;

S6, the new HSV image obtained is converted into RGB image, and after carrying out white balance process, obtains the image after strengthening.

Carry out the step of white balance process described in step S6, comprising:

S61, the color average calculating R, G, B triple channel obtaining the RGB image after transforming, and total average of RGB channel is calculated according to following formula:

K_ave=(R_ave+G_ave+B_ave)/3

In upper formula, K_aveRepresent total average of RGB channel, R_ave��G_aveB_aveRepresent the color average of R, G, B triple channel respectively;

S62, judging following formula and whether set up, if then directly terminating, not carrying out white balance process, the expression otherwise following formula is false $\frac{max(R_{ave},G_{ave},B_{ave})-min(R_{ave},G_{ave},B_{ave})}{K_{ave}}\&GreaterEqual;0.1,$ Perform step S63:

$\frac{max(R_{ave},G_{ave},B_{ave})-min(R_{ave},G_{ave},B_{ave})}{K_{ave}}<0.1$

S63, according to following formula, R, G, B color component is carried out white balance process:

$\left\{\begin{array}{c}{R}_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/R_{ave})\&times;R_{channel}\\ G_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/G_{ave})\&times;G_{channel}\\ B_{\mathrm{mod}}=\left(max\right(R_{ave},G_{ave},B_{ave})/B_{ave})\&times;B_{channel}\end{array}\right.$

In upper formula, R_mod��G_mod��B_modRepresent the color value of R, G, B triple channel after white balance process respectively, R_channel��G_channel��B_channelRepresent the gray-scale value of the front R of white balance process, G, channel B respectively.

Fig. 2 gathers the low illumination image obtained, and adopts the present embodiment that the low illumination image of Fig. 2 is carried out enhancement process, and as shown in Figure 4, the image detail after enhancing is high-visible, and color of image is normal for the result obtained. And Fig. 3 is the result obtained after adopting traditional MSRCR algorithm that the image of Fig. 2 carries out enhancement process, by the contrast of Fig. 3 and Fig. 4, present method is relative to traditional algorithm that image directly carries out enhancement process under RGB color, reinforced effects is good, and substantially can not produce color loss, more meet the visual signature of people's eye. Therefore, the present invention can greatly promote the sharpness of low illumination image, and details can be reproduced, and present method suitability is strong, robustness height.

It is more than that the better enforcement to the present invention has carried out concrete explanation, but the invention is not limited to described embodiment, those of ordinary skill in the art also can make all equivalent variations or replacement under the prerequisite not running counter to spirit of the present invention, and these equivalent modification or replacement are all included in the application's claim limited range.

Claims (8)

1. the Enhancement Method to low illumination image, it is characterised in that, comprising:

S1, obtain pending low illumination image after, it is transformed into hsv color space from RGB color, obtains chromatic(ity)component, saturation ratio component and brightness component;

S2, based on Retinex algorithm, alternately minimization method is adopted brightness component to be decomposed into reflection component and irradiates component;

S3, respectively to after irradiating component and reflection component carrying out enhancement process, synthesis strengthen after brightness component;

S4, saturation ratio component is done self-adaptative adjustment process after obtain strengthen after saturation ratio component;

S5, by chromatic(ity)component, strengthen after brightness component and strengthen after saturation ratio component synthesize new HSV image;

S6, the new HSV image obtained is converted into RGB image, and after carrying out white balance process, obtains the image after strengthening.

2. a kind of Enhancement Method to low illumination image according to claim 1, it is characterised in that, described step S2, it is specially:

Build following objective function, and obtain optimum irradiation component and reflection component according to this objective function:

$\begin{array}{ccc}\underset{R,L}{\arg min}\left|\right|RL-V|{|}_2^2+\mathrm{\&alpha;}||\&dtri;L|{|}_2^2+\mathrm{\&beta;}\left|\right|\&dtri;R|{|}_0+\mathrm{\&lambda;}||L-I_{bright}|{|}_2^2& s.t.& 0\&le;R\&le;1,I\&le;L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component, $I_{bright}=\underset{C\&Element;(R,G,B)}{max}\left(\underset{y\&Element;\mathrm{\&Omega;}}{max}\right(I_C\left(y\right)\left)\right)$ Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor.

3. a kind of Enhancement Method to low illumination image according to claim 2, it is characterised in that, described step S2, comprising:

S21, build following objective function:

$\begin{array}{ccc}\underset{R,L}{\arg min}\left|\right|RL-V|{|}_2^2+\mathrm{\&alpha;}||\&dtri;L|{|}_2^2+\mathrm{\&beta;}\left|\right|\&dtri;R|{|}_0+\mathrm{\&lambda;}||L-I_{bright}|{|}_2^2& s.t.& 0\&le;R\&le;1,I\&le;L\end{array}$

In upper formula, V represents brightness component, and L represents irradiation component, and R represents reflection component, $I_{bright}=\underset{C\&Element;(R,G,B)}{max}\left(\underset{y\&Element;\mathrm{\&Omega;}}{max}\right(I_C\left(y\right)\left)\right)$ Represent through the filtered low illumination image of maximum value, I_CY () represents low illumination image, �� represents filtering window, and ��, ��, �� are default weighting factor;

S22, structureEquivalent expression: $\left|\right|\&dtri;R|{|}_0=C\left(R\right)=\#\{p\left|\right|\&part;{\mathrm{xR}}_p|+|\&part;{\mathrm{yR}}_p|\&NotEqual;0\},$ Wherein p represents and satisfies conditionThe number of pixel;

S23, obtain according to following formulaSubsidiary variable:

$(h_p,v_p)=\left\{\begin{array}{ccc}(0,0)& if& {(\&part;{\mathrm{xR}}_p)}^2+{(\&part;{\mathrm{yR}}_p)}^2\&le;\mathrm{\&beta;}/\mathrm{\&gamma;}\\ (\&part;{\mathrm{xR}}_p,\&part;{\mathrm{yR}}_p)& if& others\end{array}\right.$

Wherein, h_pRepresentSubsidiary variable, v_pRepresentSubsidiary variable, (h_p,v_p) meet $\underset{h_p,v_p}{min}\{\underset{p}{\&Sigma;}({\left(\&part;{\mathrm{xR}}_p-h_p\right)}^2+{\left(\&part;{\mathrm{yR}}_p-v_p\right)}^2)+\frac{\mathrm{\&beta;}}{{\mathrm{\&gamma;}}_i}C(h,v)\},$ ��_iRepresent iteration variable, ��₁=2 ��=0.05 and ��_i+1=2* ��_i;

S24, using brightness component by gauss low frequency filter filtered value as irradiation component initial value;

S25, the reflection component obtaining optimum according to objective function are as follows:

$R=f^{-1}\frac{f(V/L_0)+{\mathrm{\&gamma;}}_i(f{\left(\&part;x\right)}^{*}\&times;f(h)+f{\left(\&part;y\right)}^{*}\&times;f(v\left)\right)}{f\left(1\right)+{\mathrm{\&gamma;}}_i(f{(\&part;x)}^{*}\&times;f(\&part;x)+f{(\&part;y)}^{*}\&times;f(\&part;y))}$

Wherein, L₀Representing the initial value irradiating component, f represents FFT, f^-1Represent FFT inverse transformation, f ()^*Hetero conjugation after expression FFT, h representsSubsidiary variable;

S26, according to following formula to reflection variable correct: R=min (max (R, 0), 1);

The value of the reflection component after S27, maintenance rectification is constant, obtains optimum irradiation component according to objective function as follows:

$L=f^{-1}\frac{f(V/R+{\mathrm{\&lambda;I}}_{bright})}{f(1+\mathrm{\&lambda;})+\mathrm{\&alpha;}(f{\left(\&part;x\right)}^{*}\&times;f(\&part;x)+f{\left(\&part;y\right)}^{*}\&times;f(\&part;y\left)\right)}$

S28, according to following formula to irradiation variable correct: L=max (L, V);

S29, judge iteration variable ��_iWhether it is greater than predetermined threshold value, if then terminating, otherwise after iteration number of times is added 1, using the irradiation component after rectification as the initial value irradiating component, returning and performing step S25.

4. a kind of Enhancement Method to low illumination image according to claim 1, it is characterised in that, described step S3, comprising:

S31, irradiation component is carried out gamma correction, and after carrying out adaptive histogram equalization process, obtain the irradiation component after strengthening;

S32, reflection component is carried out obtaining after part filter operation the reflection component after strengthening;

S33, by strengthen after irradiation component and strengthen after reflection component synthesis strengthen after brightness component.

5. a kind of Enhancement Method to low illumination image according to claim 4, it is characterised in that, described in described step S31, irradiation component is carried out the step of gamma correction, it is specially:

According to following formula, irradiation component is carried out gamma correction:

$L_g=L^{\frac{1}{\mathrm{\&gamma;}}},$ ��=2.2

In upper formula, L represents irradiation component, L_gRepresent the irradiation component after gamma correction.

6. a kind of Enhancement Method to low illumination image according to claim 4, it is characterised in that, described step S32, it is specially:

According to following formula, after reflection component is carried out part filter operation, obtain the reflection component after strengthening:

$R_F(x,y)=\frac{1}{K}\&times;\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}{w}_s(i,j)w_r(i,j)R(i,j)$

In upper formula, R_F(x, y) representing that the reflection after the reflection component of pixel (x, y) is carried out part filter operation divides value, �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window, R (i, j) pixel (i is represented, j) reflection divides value, and K represents normalization method constant, w_s(i, j) represents the spatial domain weight of pixel (i, j), w_r(i, j) represents the codomain weight of pixel (i, j), and K, w_s(i,j)��w_r(i, j) meets the following conditions:

$\left\{\begin{array}{c}K=\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}{w}_s(i,j)w_r(i,j)\\ w_s(i,j)=\exp (-\frac{{\left|i-x\right|}^2+{\left|j-y\right|}^2}{2{\mathrm{\&sigma;}}_s^2})\\ w_r(i,j)=\underset{m=x}{\overset{i-1}{\&Sigma;}}{w}_r(m,y;m+1,y)\&times;\underset{n=y}{\overset{j-1}{\&Sigma;}}{w}_r(i,n;i,n+1)=\exp (-\frac{\underset{m=x}{\overset{i-1}{\&Sigma;}}{\left|R_{m,y}-R_{m+1,y}\right|}^2\&times;\underset{n=y}{\overset{j-1}{\&Sigma;}}{\left|R_{i,n}-R_{i,n+1}\right|}^2}{2{\mathrm{\&sigma;}}_r^2})\end{array}\right.$

Wherein, ��_SRepresent spatial domain standard deviation, ��_rRepresent codomain standard deviation, R_m,yRepresent the intensity level of pixel (m, y), R_i,nRepresent the intensity level of pixel (i, n).

7. a kind of Enhancement Method to low illumination image according to claim 1, it is characterised in that, described step S4, it is specially:

According to following formula, saturation ratio component is done after self-adaptative adjustment processes and obtains the saturation ratio component after strengthening:

S'=S+t �� (V'-V) �� ��

In upper formula, S' represents the saturation ratio component after enhancing, and S represents saturation ratio component, and t is constant, and V represents brightness component, and V' represents the brightness component after enhancing, and �� represents regulation coefficient, and for each pixel (x, y), regulation coefficient is:

$\mathrm{\&mu;}(x,y)=\underset{(i,j)\&Element;\mathrm{\&Omega;}}{\&Sigma;}\left(\right|\left(V\right(i,j)-\stackrel{\&OverBar;}{V_{\mathrm{\&Omega;}}}(i,j\left)\right)|\&times;|\left(S\right(i,j)-\stackrel{\&OverBar;}{S_{\mathrm{\&Omega;}}}(i,j\left)\right)\left|\right)/\sqrt{{\mathrm{\&delta;}}_V(x,y)\&times;{\mathrm{\&delta;}}_S(x,y)}$

Wherein, (x, y) represents pixel position, and �� represents with pixel (x, the neighborhood window of the n �� n centered by y), (i, j) represents the pixel coordinate in this neighborhood window ��, V (i, j) brightness value of pixel (i, j) is representedRepresenting the luminance mean value in neighborhood window ��, S (i, j) represents the intensity value of pixel (i, j),Represent the saturation ratio average in neighborhood window ��, ��_V(x, y) represents the brightness variance of pixel (x, y), ��_S(x, y) represents the saturation ratio variance of (x, the y) of pixel.

8. a kind of Enhancement Method to low illumination image according to claim 1, it is characterised in that, carry out the step of white balance process described in described step S6, comprising:

S61, the color average calculating R, G, B triple channel obtaining the RGB image after transforming, and total average of RGB channel is calculated according to following formula:

K_ave=(R_ave+G_ave+B_ave)/3

In upper formula, K_aveRepresent total average of RGB channel, R_ave��G_aveB_aveRepresent the color average of R, G, B triple channel respectively;

S62, judge following formula and whether set up, if then directly terminating, otherwise, perform step S63:

$\frac{max(R_{ave},G_{ave},B_{ave})-min(R_{ave},G_{ave},B_{ave})}{K_{ave}}<0.1$

S63, according to following formula, R, G, B color component is carried out white balance process:

$\left\{\begin{array}{c}{R}_{\mathrm{mod}}=\left(\max \right(R_{ave},G_{ave},B_{ave})/R_{ave})\&times;R_{channel}\\ G_{\mathrm{mod}}=\left(\max \right(R_{ave},G_{ave},B_{ave})/R_{ave})\&times;G_{channel}\\ B_{\mathrm{mod}}=\left(\max \right(R_{ave},G_{ave},B_{ave})/B_{ave})\&times;B_{channel}\end{array}\right.$

In upper formula, R_mod��G_mod��B_modRepresent the color value of R, G, B triple channel after white balance process respectively, R_channel��G_channel��B_channelRepresent the gray-scale value of the front R of white balance process, G, channel B respectively.